Conditional recall and the frequency effect in the serial recall task: an examination of...

transcript

Conditional recall and the frequency effect in the serial recalltask: an examination of item-to-item associativity

Leonie M. Miller & Steven Roodenrys

Published online: 13 June 2012# Psychonomic Society, Inc. 2012

Abstract The frequency effect in short-term serial recall isinfluenced by the composition of lists. In pure lists, a robustadvantage in the recall of high-frequency (HF) words isobserved, yet in alternating mixed lists, HF and low-frequency (LF) words are recalled equally well. It has beenargued that the preexisting associations between all listitems determine a single, global level of supportive activa-tion that assists item recall. Preexisting associations betweenitems are assumed to be a function of language co-occurrence; HF–HF associations are high, LF–LF associa-tions are low, and mixed associations are intermediate inactivation strength. This account, however, is based onresults when alternating lists with equal numbers of HFand LF words were used. It is possible that directionalassociation between adjacent list items is responsible forthe recall patterns reported. In the present experiment, therecall of three forms of mixed lists—those with equal numb-ers of HF and LF items and pure lists—was examined to testthe extent to which item-to-item associations are present inserial recall. Furthermore, conditional probabilities wereused to examine more closely the evidence for a contribu-tion, since correct-in-position scoring may mask recall thatis dependent on the recall of prior items. The results suggestthat an item-to-item effect is clearly present for early but notlate list items, and they implicate an additional factor, per-haps the availability of resources at output, in the recall oflate list items.

Keywords Short termmemory . Language production .

Recall

In the context of immediate serial recall, the frequency effectis complex and difficult to explain in straightforward terms(e.g., Hulme, Stuart, Brown, & Morin, 2003; Morin, Poirier,Fortin, & Hulme, 2006). While tasks involving pure lists ofeither only high-frequency (HF) or only low-frequency (LF)words have shown a clear advantage for HF words acrossmany replications (Allen & Hulme, 2006; Hulme et al., 1997;Hulme et al., 2003; Morin et al., 2006; Poirier & Saint-Aubin,1996; Stuart & Hulme, 2000; Roodenrys, Hulme, Lethbridge,Hinton, & Nimmo, 2002; Roodenrys & Quinlan, 2000; Saint-Aubin & LeBlanc, 2005; Saint-Aubin & Poirier, 2005;Watkins & Watkins, 1977), the results of experiments usingmixed lists, where HF and LF words appear together on thesame trial, indicate that under some conditions, LF items canbe recalled as well as HF words (Hulme et al., 2003; Morin etal., 2006; Saint-Aubin & LeBlanc, 2005).

Early explanations of the frequency effect in serial recall,motivated by the results from experiments using pure lists,focused on item-specific differences between HF and LFwords, consistent with the prevailing explanation of verbalshort-term memory (STM) phenomena. For example, thephonological loop (Baddeley & Hitch, 1974) was character-ized as a serially ordered, speech-based system that wasresponsible for performance on memory span and serial recalltasks (Baddeley, 1986). It comprised a phonological short-term store (STS) that retained the short-term traces of itemsand a subvocal rehearsal mechanism. Traces in the STS weresubject to passive decay and consequent degradation unlessrefreshed by rehearsal in the loop. Wright (1979) demonstrat-ed articulation rate differences between HF and LF words; HFitems of the same length are articulated faster than LF words.Accordingly, the frequency effect will manifest from rehearsalrate differences between HF and LF words, since greaterrehearsal efficiency leads to superior retention of the short-term traces on which recall is reliant.

L. M. Miller (*) : S. RoodenrysSchool of Psychology, University of Wollongong,Wollongong, NSW 2522, Australiae-mail: leoniem@uow.edu.au

Mem Cogn (2012) 40:1246–1256DOI 10.3758/s13421-012-0221-5

However, when the effect in pure lists was tested underconditions preventing rehearsal (Gregg, Freedman, &Smith, 1989; Tehan & Humphreys, 1988) or when differ-ences in articulation rate were statistically accounted for(Hulme et al., 1997; Hulme et al., 2003; Stuart & Hulme,2000), a frequency effect persisted, implicating a secondsource of effect. Hulme et al. (1997) nominated phonolog-ical long-term memory (LTM) as this source; a second-stageprocess utilizing available long-term phonological traces,redintegration (Schweickert, 1993), would restore degradedshort-term traces through pattern matching and completion.It was argued that such traces were more accessible for HFthan for LF words and, consequently, HF items were morelikely to be successfully redintegrated. Notably, in this ac-count, differences in recall were attributed to another item-specific property—namely, access to phonological long-termrepresentations.

More recently, the notion that item-specific propertiesform the basis of the frequency effect was challenged by aseries of experiments demonstrating list-dependent influen-ces on the recall of HF and LF words (Hulme et al., 2003;Morin et al., 2006; Saint-Aubin & LeBlanc, 2005; Stuart &Hulme, 2000). Stuart and Hulme investigated whether themanipulation of preexperimental association between HFand LF items affected recall performance. Their study didnot “mix” HF and LF items within the same list but com-pared recall for lists of items that had been familiarized bypairwise association during a training period with that forlists of items familiarized individually. Stuart and Hulmeobserved a benefit to the recall of pure LF lists after famil-iarization that did not occur for pure HF lists. They arguedthat familiarization per se did not result in better recall,while LF lists constructed from the same subset of items—that is, lists containing pairwise familiarized items—wererecalled better than the alternating lists where adjacent itemswere not from the same familiarization pool.1 Consequently,the authors argued that the frequency effect could beexplained entirely in terms of associative links betweenitems. Instead of the frequency effect being driven by theaccessibility of individual item representations in LTM, theysuggested that it might be an outcome of the preexperimen-tal associations formed from the co-occurrence of items innatural language (Deese, 1960). The representations ofitems would form a “mutually supporting network of itemnodes” (p. 801) facilitating the accessibility of each item’slong-term trace in the retrieval process; that is, the associa-tive links would mutually excite connected list members and

determine the accessibility of these LTM representations(Saint-Aubin & Poirier, 2005).

Hulme et al. (2003) showed that in alternating lists of HFand LF words, the frequency effect is eliminated (Experi-ment 1) or much reduced (Experiment 2). The pattern ofitem errors mirrored the serial recall performance across thelist types; the pure list advantage for HF items was nullified(Experiment 1) or unreliably reversed (Experiment 2) inmixed lists. It was argued that such outcomes were furthersupport for the influence of preexperimental associationsbetween list items in redintegration (Stuart & Hulme,2000). Furthermore, Morin et al. (2006) identified that theabolition of the frequency effect with alternating lists wasnot an outcome of task awareness, since it occurred underboth incidental and intentional learning conditions. Accord-ingly, Morin et al. (2006) argued that the STM processesresponsible for the nature of the frequency effect are notconsciously mediated and nominated Hulme et al.’s (2003;Stuart & Hulme, 2000) redintegration hypothesis as thepreferred explanation.

Nonetheless, Hulme et al. (2003) referred to a model ofepisodic memory, the temporal context model (TCM;Howard & Kahana, 2002a, 2002b), as an alternate concep-tual framework suggesting how associative effects mightunderlie the frequency effect. In this model, item recall isdependent upon the reinstantiation of the temporal contextat encoding, and recall of an item then acts as a cue toestablish the temporal context for the next item to berecalled, and so on. The encoded temporal contexts aredetermined by a slowly evolving context state that is com-bined with former contexts in which an item has beenencoded. Hence, items presented close together in time willhave some overlap in their contextual states, and historicalco-occurrence of items, as in the case of strong semanticassociates, will enhance the degree of contextual overlap. Ac-cordingly, TCM states that interitem associations drive theforward order of recall and the stronger the semantic associa-tion between successive items, the greater the likelihood thatthe first item in a pair will facilitate the retrieval of the second.Furthermore, it suggests that effects of interitem semanticassociation should be localized in the associative strengths ofconsecutive pairs of list items.

This contrasts with the position taken by Hulme et al.(2003; Stuart & Hulme, 2000), where the interitem associa-tions of all list items were seen to operate in a mutuallysupporting and nondirectional manner, facilitating their re-trieval. The alternating lists used by Hulme et al. (2003) didnot provide an adequate test of these positions, sincefrequency-wise alternation would yield intermediate levelsof associative strength between consecutive pairs of itemsacross the list (HF item to LF item or vice versa) and inter-mediate association strength at the list level (since each mixedlist contained three HF and three LF items). The question of

1 The manipulation of familiarity used by Stuart and Hulme (2000) hasbeen called into question by Saint-Aubin and Poirier (2005), whoproposed that familiarity was confounded with set size in this experi-ment. They found a comparable effect of familiarity when items arefamiliarized alone and argued against associativity as the mechanismresponsible for the frequency effect.

Mem Cogn (2012) 40:1246–1256 1247

whether associative effects are directional or nondirectionalremained unresolved.

Recently, Miller and Roodenrys (2012) presented theresults of experiments using a half-listmanipulation originallyproposed by Hulme et al. (2003) as a clear test of the direc-tionality of associative effects in lists of mixed frequency. Halflists are constructed from the halves of pure lists so that HFand LF items are presented in sequence (HHHLLL andLLLHHH) and match the list composition of alternating listswith three HF and three LF items each. A sample was testedon the recall of pure and half lists; it was observed that halflists mimicked the performance with pure lists in the first halfof the list, while recall for half lists in the second half did notdiffer statistically between list types and was intermediatewith respect to performance with pure lists (except for apossible difference in the recency position). In descriptiveterms, HHHLLL lists behaved like HF pure lists, andLLLHHH lists matched recall for pure LF lists for the firstthree serial positions. Recall performance for half lists inter-sected between positions 4 and 5, well after the change in listcomposition (see Fig. 1 for a replication of this feature).Clearly, these results are not consistent with a nondirectionalassociative redintegration explanation, since it predicts thatserial position curves for half lists should be the same as thosefound with alternating lists (Hulme et al., 2003). In contrast, a

directional mechanismmight predict an abrupt shift in recall atposition 4 in the half lists, but this was not observed. Accord-ingly, it was proposed that recall was unlikely to be dictated bypurely directional associative influences.

However, the lag in the crossover in half-list recall with thechange in list composition is not necessarily inconsistent withthe notion of interitem association that operates between suc-cessive items. While changes in interitem association wouldbe expected to alter the rate of change in correct recall betweenitems, the absolute difference between serial position curvesprior to a change in list composition might dictate how farbeyond a change in list composition the crossover actuallyoccurs. Furthermore, when absolute recall levels are similar,as occurs at crossover points, the capacity of item-to-itemassociativity to generate significant effects between lists insubsequent positions might be limited.

The present work sought to further clarify the extent towhich directional associations influence the recall of list items,in two ways. First, a between-subjects design comparing thewithin-subjects serial recall of two list types included pure,alternating, and half lists, and a new format—sequence lists(HHLLLH or LLHHHL)—was used to examine in greaterdetail the serial recall behavior in lists where a sequence ofitems of the same frequency type occurred after an earlytransition between HF and LFwords. The sequence list format

1 2 3 4 5 6

HF pure list

LF pure list

1 2 3 4 5 6

H-L alternating list

L-H alternating list

1 2 3 4 5 6

Serial position

HHHLLL

LLLHHH

1 2 3 4 5 6

Serial position

HHLLLH

LLHHHL

Fig. 1 Serial recall of wordsas a function of frequencyand list composition. Panelsrepresent the between-subjectsconditions

1248 Mem Cogn (2012) 40:1246–1256

is a variation on the half-list composition, where the sequenceof same-frequency items in the second half of the list isbrought forward by one serial position and the displaced itemtype from the first half sequence is moved to the last serialposition. This format maintains the same composition as theother mixed lists formats (three HF and three LF items) andhas the added benefit of deconfounding directional associativeeffects with any other influences arising from the coincidenceof the last item in sequence with the recency position, asoccurs for half lists.

Second, this experiment sought to explore more closelyrecall at points of transition within mixed lists. If associativityis bidirectionally equivalent, and if the recall of an item is afunction of the interitem associativity that it shares with theprevious item in the list (Howard & Kahana, 2002b), then thelikelihood it will be recalled, given recall of the previous item,should be the same for HF-to-LF and LF-to-HF transitions.The standard serial recall analysis masks these sensitivities,since all instances of successful serial recall at each serialposition are reported. Accordingly, in this study, a novelapproach to recall scoring was adopted where recall at eachposition was examined when conditionalized upon successfulrecall of the previous item.

Method

Participants

A total of 103 undergraduate University of Wollongongstudents participated in the experiment for course credit.The data from 7 participants were excluded from analysisbecause they were not native Australian English speakers(6), or were visually impaired (1). The remaining 96 partic-ipants were allocated to one of four conditions. The pure listparticipants (21 female, 3 male) had a mean age of 23.8 years(SD 0 8.6 years), the alternating list participants (18 female,6 male) had a mean age of 22.2 years (SD 0 8.0 years), thehalf-list participants (22 female, 2 male) had a mean age of22.8 years (SD 0 5.8 years), and the sequence list partic-ipants (19 female, 5 male) had a mean age of 21.0 years(SD 0 3.8 years).

Materials

The stimulus sets were those used by Miller and Roodenrys(2012). The two sets of 96 CVCwords selected on the basis offrequency ratings (instances per million words of text) weredrawn from the Celex database (Baayen, Piepenbrock, & VanRijan, 1993). These ratings combined database entries for thesame orthography and the same word identification number,so that, for example, the frequency counts for bird and birds

were tallied. The counts of any homophones were also includ-ed in the frequency ratings. The mean log10 frequency ratingsof the sets were, for LF,M 0 0.70 (SD 0 0.37) and for HF,M 0

2.21 (SD 0 0.27), and the words sets were found to differsignificantly on raw frequency ratings, Mann–Whitney U 0

0.00, p < .001. Sets were matched on concreteness (MRCdatabase; Coltheart, 1981), using a weighted average con-creteness value for homophone items calculated with therelative frequencies as weights, U 0 4,495.50, p 0 .770, andthe number of phonological neighbors (derived from theCelex database), U 0 4,560.50, p 0 .902. Furthermore, thenumber of items that were phonological neighbors of otheritems in the set did not differ between sets, U 0 4,555.50,p 0 .890.

Phonological similarity was measured using an ExcelVisual Basic program that compares phoneme feature sim-ilarity for onset, vowel, and coda segments of stimuli (PSI-METRICA; Mueller, Seymour, Kieras, & Meyer, 2003).Pairwise comparisons are made for all stimuli of a set, andthe dissimilarity profile is determined as the average of eachcluster measure across all pairwise combinations. None ofthe resultant phonological similarity measures differed sig-nificantly between sets (onset, U 0 4,117.00, p 0 .202;nucleus, U 0 4,398.00, p 0 .585; coda, U 0 4,365.00,p 0 .527).

Vowel quality (categorized as short vowels, long vowels,and diphthongs) in each set was measured using the DeCaraand Goswami (2002) database. A 3 × 2 χ2 test for indepen-dence identified that a marginal difference in vowel qualitywas present between HF and LF sets, χ2(2) 0 4.78, p 0 .091.However, since there were more words with short vowelsand fewer words with long vowels in the LF set, any effectfavoring HF words in serial recall cannot be attributed todifferences in vowel quality.

Recent work has highlighted the possibility that co-articulatory fluency—namely, the ease with which articulatorytransitions between items are performed—can influence serialrecall performance (Murray & Jones, 2002; Woodward,Macken, & Jones, 2008). More specifically, this accountoffers a rival explanation of the frequency effect—namely,that the transitions between HF items are more easily negoti-ated than transitions between LF items (Woodward et al.,2008). Accordingly, the stimulus sets were examined in rela-tion to the coarticulation characteristics (manner and place ofarticulation and voicing) of all coda–onset combinations(9,120 boundaries per word set). This analysis suggested thatdifferences in articulatory complexity between stimulus setswere unlikely to contribute markedly to a frequency effect inserial recall. The proportions of word boundaries with thesame manner of articulation were similar for HF and LF sets(.264 vs. .279). Transitions involving the same place of artic-ulation were similar for HF and LF sets (.325 vs. .319).Although boundaries likely to be easily assimilated (those

Mem Cogn (2012) 40:1246–1256 1249

with easily coarticulated coda–onset consonants—e.g.,phone → ball) were greater for HF than for LF words(.131 vs. .089), instances of minor place changes were fewerfor HF than for LF words (.234 vs. .309). Furthermore, moreinstances of word boundaries with major place changes inarticulation occurred in the HF set (.310 vs. .285). HF andLF sets possessed similar proportions of cases with boundariesinvolving no change of voicing (.509 vs. .516).

A final check of possible differences in coda–onset coar-ticulation in the word sets used transitional probabilities ofbiphone frequency in language. These measures quantifyhow often two phonemes occur in order in natural language(Toro, Nespor, Mehler, & Bonatti, 2008). It is argued thattransitional probabilities can be used as a proxy to estimatecoarticulation fluency between words. The biphone frequen-cy database of Frankish (unpublished) that is sourced fromCelex database information provided transitional probabili-ties of all coda–onset combinations in each set. Setwisedistributions were found to be positively skewed, so thesedata were subjected to a square root transformation. Nosignificant difference between HF and LF sets (MDiff 0

.002), t(18238) 0 1.62, p 0 .105, was found on the trans-formed data sets, suggesting that the frequency with whichtransitions were encountered in either set was similar interms of biphone frequency.

The stimuli were digitally recorded by a female nativeAustralian English speaker and were converted to sound filesusing the ProTools LE software on a G4 Macintosh computer.Experimental sessions were conducted on an IBM-compatiblePC that ran prepared script files loaded into SuperLab v. 2.0.4.Sound files were amplified using an external speaker that wasattached to the PC.

Script files contained 64 six-word trials and tested one offour list format conditions: pure frequency, alternating, half,or sequence lists. Each item of the HF and LF word sets waspresented twice within script files, once within each of thelist types of the list format. Allocation of items to serialpositions was random within the constraints of each listformat, and the presentation of individual trial types wasrandom.

Procedure

Participants were assigned to each of the list format con-ditions on a rotating basis, according to the order of testing.All participants were tested individually. The total time tocomplete the experiments was approximately 45 min. Initi-ation of each trial occurred when the participant pressed thespace bar of the keyboard. The program would then play thesound files of each word in the trial at a rate of one word persecond. After the presentation of the sixth word, a recallprompt (“?????”) appeared on the screen, indicating that theparticipant should commence recall. Spoken recall was

according to strict serial recall criteria—namely, (1) wordswere recalled in order of presentation; (2) if a word couldnot be recalled, the participant would indicate by saying“blank”; and (3) previous items were not to be recalled afterparticipants moved on to successive items in the list. Theexperimental program presented four practice trials that wereused to familiarize participants with the task requirementsbefore the commencement of the experimental phase.

Results

The data were scored according to a strict serial recall criteri-on; that is, a recalled itemwas considered to be correct if it wasrecalled in its presentation serial position. The mean numberof correctly recalled items by serial position and condition isshown in Fig. 1, where the distinct effects of list compositionare readily observed. The patterns for the pure, alternating,and half-list formats replicated those obtained in previousexperiments (Hulme et al., 2003; Miller & Roodenrys,2012), while the novel sequence format was consistent withthe half-list condition, in that the convergence of the curvesoccurred at the point of change of item type. The means forformats collapsed across list types suggested that the variationin the overall level of performance for each condition wassmall; descriptively, the items in pure lists were recalled theleast well (M 0 .535), followed by those in alternating lists(M 0 .541) and then half lists (M 0 .559), while the items in thesequence lists were recalled the most (M 0 .583). However,since list format was the between-groups variable in the ex-periment, this variation may be participant related. The overallmeans for the list types of each format revealed that listsbeginning with HF words were recalled better than lists be-ginning with LF words: pure lists, HF (M 0 .634 ) versus LF(M 0 .436); alternating lists, HLHLHL (M 0 .556) versusLHLHLH (M 0 .526); half lists, HHHLLL (M 0 .576) versusLLLHHH (M 0 .541); and sequence lists, HHLLLH(M 0 .601) versus LLHHHL (M 0 .566).

In summary, it would appear that HF items in mixed listsbeginning with LF words were recalled better than theircounterparts in only a handful of serial positions. Further-more, the differences where performance was superior inthese lists did not compensate for the advantage that listsbeginning with HF words realized in the other serialpositions.

Serial recall

An alpha level of .05 was applied to the statistical testsperformed. The data were subjected to a 4 × 2 × 6 mixedanalysis of variance where list format (pure, alternating,half, and sequence) was the between-subjects factor and listtype (lists commencing with HF or LF words) and serial

1250 Mem Cogn (2012) 40:1246–1256

position (1–6) were the within-subjects factors. This analy-sis revealed a main effect of list type, F(1, 92) 0 138.30, p <.001, ηp

2 0 .601, confirming that lists beginning with HFwords were recalled better than lists beginning with LFwords, and a main effect of serial position, F(5, 460) 0

363.07, p < .001, ηp2 0 .798, but the effect of format was

nonsignificant, F(3, 92) 0 0.82, p 0 .480, ηp2 0 .026. Thus,

differences in mean performances between formats were notreliable. The list type × list format interaction was signifi-cant, F(3, 92) 0 42.80, p < .001, ηp

2 0 .583, reflectingprimarily the greater difference in recall of pure list typesin comparison with differences between list types in theother formats. The list type × serial position interactionwas also significant, F(5, 460) 0 9.96, p < .001, ηp

.098, driven by the asymmetries in the list types for the halfand sequence formats. However, the interaction of formatand serial position was nonsignificant, F(15, 460) 0 1.07,p 0 .382, ηp

2 0 .034, identifying that there was no differenceacross serial position in the average recall between formats.This result was qualified by a significant three-way interac-tion (format × list type × serial position), F(15, 460) 0 9.70,p < .001, ηp

2 0 .240, which demonstrated that when perfor-mance was considered by list type, differences in serialrecall curves between lists beginning with HF items andthose beginning with LF items varied across the fourformats.

This interaction was explored further by analyzing the datafor each format separately as four 2 × 6 repeated measuresanalyses of variance. Pure lists contained a significant maineffect of list type, F(1, 23) 0 199.19, p < .001, ηp

2 0 .896,replicating the standard frequency effect, and a significantmain effect of serial position, F(5, 115) 0 111.23, p < .001,ηp

2 0 .829. The list type × serial position interaction was alsosignificant, F(5, 115) 0 5.52, p < .001, ηp

2 0 .194, indicatingthat the effect increased over the first serial positions. Toresolve whether the interaction was a result of a ceiling effectoperating on the first position for HF lists, the last five serialpositions were reanalyzed. The main effects were once againsignificant [list type, F(1, 23) 0 219.10, p < .001, ηp

2 0 .905;serial position, F(4, 92) 0 61.80, p < .001, ηp

2 0 .729],but the interaction was nonsignificant, F(4, 92) 0 0.13, p 0

.971, ηp2 0 .005, supporting a ceiling effect interpretation of

the interaction.A 2 × 6 repeated measures analysis of variance on the

alternating list data produced a significant main effect of listtype, F(1, 23) 0 5.63, p 0 .026, ηp

2 0 .197. Thus, lists thatbegan with an HF item were recalled modestly but reliably,better than those beginning with an LF item. The main effectof serial position was significant, F(5, 115) 0 82.82, p <.001, ηp

2 0 .948, as was the list type × serial positioninteraction, F(5, 115) 0 6.30, p < .001, ηp

2 0 .612. Thepresence of an interaction was due presumably to the subtlesawtooth pattern present in both list types, indicating the

possibility of small item-specific effects (Saint-Aubin &LeBlanc, 2005).

The equivalent analysis of the half-list format data resultedin main effects of list type, F(1, 23) 0 11.79, p 0 .002, ηp

2 0.339, and serial position, F(5, 115) 0 127.56, p < .001, ηp

.847. Furthermore, the list type × serial position interactionwas significant, F(5, 115) 0 16.91, p < .001, ηp

2 0 .424,highlighting the superiority of lists beginning with HF words.Bonferroni-adjusted simple effects identified a significant fre-quency effect for positions 1, 2, 3, and 6. Therefore, at thepoint of change in item frequency (the fourth serial position),performance converged, was similar for the fifth position, andthen diverged at the recency position, where HF items wererecalled better than LF items.

Lastly, a within-subjects analysis of the sequence list datarevealed main effects of list type, F(1, 23) 0 6.51, p 0 .018,ηp

2 0 .221, and serial position, F(5, 115) 0 63.13, p < .001,ηp

2 0 .733. The list type × serial position interaction wasalso significant, F(5, 115) 0 9.73, p < .001, ηp

2 0 .297. Theinteraction adopted the pattern for half lists over the initialserial positions, confirming the better recall of the HHLLLHthan of the LLHHHL list type. Simple effects analysis usingBonferroni adjustment revealed that significant frequencyeffects were present in only the first two serial positions. Atthe first change in list composition (position 3), performanceconverged; however, in this condition, the difference inrecall between list types was not reliable for the remainderof the serial positions (3–6).

Using conditional probabilities to examine interitem effects

To explore the possibility that serial recall is driven, at leastin part, by some item-to-item associative mechanism, thedata in the experiment were rescored according to a condi-tional recall criterion. Should an interitem associative mech-anism influence recall, it would be expected that theconditional likelihood that an item would be recalled wouldbe the same for transitions between HF and LF items inmixed lists, regardless of the order of items, since these havebeen argued to possess interitem association of intermediatestrength (Hulme et al., 2003).

Despite concerns regarding the accuracy of dependencymeasures (see Henson, Norris, Page, & Baddeley, 1996), itwas thought, given the use of open stimulus sets in thisinstance, that the use of transitional shift probabilities, (i.e.,conditionalized probabilities based on the recall status of theprevious item only) should be sufficient to indicate depen-dency on the relationship between adjacent items. However,it is acknowledged that effects of disruption on recall of laterlist items may mask dependencies due to item-to-item asso-ciation. Accordingly, it is appropriate to remain mindful thatmeasures of this sort for later serial positions might containinfluences from a number of sources.

Mem Cogn (2012) 40:1246–1256 1251

A strict position interpretation was adopted for these data,due to the short length of the supraspan lists. That is, theproportion of instances where an item presented in serialposition j was recalled in position j on the condition that theprevious items in position j − 1 had also been recalledcorrectly was recorded. These data would identify anyinstances where interitem associations between list itemscould explicitly operate. To account for the changing sizeof the sample space when calculating the conditional prob-abilities associated with these proportions, the followingformula was used:

P j j j� 1ð Þ ¼ P j \ j� 1ð ÞP j� 1ð Þ ;

where P( j | j − 1 ) is the probability of recall of an item inposition j subject to the correct recall of the previous item j −1, P( j − 1) is the probability of correct recall of the prior item,and P( j∩ j − 1 ) is the probability of the correct recall of itemsj and j − 1 (Larsen &Marx, 1981). The term P( j − 1) becomesan estimate for the adjusted sample space. These probabilitiesare presented for each mixed list condition in Fig. 2a. The datafor the first serial position are the proportions recalled as perthe original serial recall scoring, since there is no previousevent for this position. The equivalent data for the pure listsare presented for each mixed list condition and act as anenvelope for the mixed list values.

The identification of recall events in this way allows thedata to be fractionated into continuous recall, as describedabove, and those instances where recall occurred despiterecall of the previous item being in error, termed recoveryrecall. In this situation, recall might reflect an item-specificinfluence and indicate how well recall can recover fromdisruption at output for serial positions 2–6. The conditionalprobabilities for the recovery data are given in Fig. 2b. Thecontinuous and recovery recall data sets were examinedseparately.

Simple effects on the conditionalized data were con-ducted to determine whether they conformed to the patternsthat directional interitem associativity would anticipate.With respect to continuous recall, frequency effects wouldbe expected in serial positions where corresponding sequen-ces of HF and LF items were presented in a condition, whileno difference in effect would occur at transition points inthese lists. Specifically, sequences of HF items should berecalled better than sequences of LF words because HFitems have stronger preexperimental association, and HF-to-LF and LF-to-HF transitions in lists should result in thesame level of recall. Therefore, it would be predicted thatpositions 2, 3, 5, and 6 in half lists and positions 2, 4, and 5in sequence lists would produce differences. In contrast, nodifference between the conditional probabilities of continu-ous recall should exist for all positions in alternating lists

(2–6), while position 4 in the half lists and positions 3 and 6in the sequence lists should also be equivalent. Additionally,under the assumption that the fractionation of continuousand recovery recall accurately separates interitem and item-specific effects, it would be expected that if item-specificeffects do not influence recall, no differences in the recoverydata should exist for any of the conditions.

The analysis found that all points of transition betweenHF and LF items, except for position 5 in the alternatinglists, produced nonsignificant differences in the conditionalprobabilities for continuous recall; the exception was foundto be a marginal result. Furthermore, significant differenceswere found for positions 2, 3, and 6 in half lists and position2 in sequence lists. In summary, one out of the eight posi-tions where no difference was predicted produced a margin-al effect, while four out of the seven positions predicted toproduce a frequency effect did so. Therefore, the continuousdata, particularly across the first four serial positions, wereconsistent with a directional interitem associativity explana-tion of serial recall.

The recovery recall identified that a significant differenceoccurred in position 3 of the sequence lists. That is, at thepoint of transition in the list, HF words were recoveredbetter than LF words after the previous item was not cor-rectly recalled. None of the other comparisons for these datareached significance, although position 3 for alternating listswas marginal. In general then, according to this analysis,recovery episodes were free of the influence of frequency.

Item analysis

Responses in each trial were classified as correct (recall ofthe item occurred in the correct serial position), an ordererror (the item belonged to another serial position in thetrial), or an item error (intrusions, omissions, and repeti-tions). The proportion of errors of each classification, foreach item type in each list type, collapsed across serialposition and participants is given in Table 1, together withthe proportion of items correct.

Order errors were conditionalized to indicate the extent towhich recalled items were positioned incorrectly, relative topresentation order (Murdock, 1976; Poirier & Saint-Aubin,1996; Saint-Aubin & Poirier, 1999). This process identifiedthat the adjusted error rate for HF items in pure lists was .127,for LF items in pure lists was .149, for HF items in alternatinglists was .119, for LF items in alternating lists was .115, for HFwords in half lists was .110, for LF words in half lists was .122,for HF words in sequence lists was .100, and finally, for LFwords in sequence lists was .099.

A 4 × 2 (list format × item type) mixed ANOVA wasperformed on the conditionalized order data and identifiednonsignificant effects of frequency F(1, 92) 0 2.35, p 0 .129,ηp

2 0 .025, and list format,F(3, 92) 0 1.86, p 0 .142, ηp2 0 .057,

1252 Mem Cogn (2012) 40:1246–1256

and a nonsignificant frequency × format interaction, F(3, 92) 01.82, p 0 .147, ηp

2 0 .056. Therefore, there was no differencebetween HF and LF stimuli in the proportion of all itemsremembered but recalled in the wrong position.

The analysis of the total item error data revealed an effect offrequency, F(1, 92) 0 254.46, p < .001, ηp

2 0 .734, and asignificant frequency × list format interaction, F(3, 92) 050.08, p < .001, ηp

2 0 .620, but the main effect of list formatwas not significant, F(3, 92) 0 0.50, p 0 .683, ηp

2 0 .016.Across conditions, HF words incurred fewer item errors thandid LF words. The interaction reflected the greater frequencyeffect for pure than for mixed lists. Bonferroni-adjusted testson each list format identified, however, that all conditionsproduced significant frequency effects [pure lists, t(23) 0

16.66, p < .001; half lists, t(23) 0 5.88, p < .001; sequencelists, t(23) 0 4.51, p < .001; alternating lists, t(23) 0 3.52,p 0 .002].

Discussion

An objective of this experiment was to explore the impact ofshifting a three-item sequence, as found in the second half of

half lists, one position forward. This arrangement served toavoid the coincidence of the third item in sequence with therecency position and to test whether frequency effects be-tween list items were possible in the second half of the listafter a transition between HF and LF items.

The analysis of correct recall for sequence lists failed tofind a significant frequency effect for the third HF and LFitems in sequence (the serial position 5). Therefore, accord-ing to these data, influences of preexperimental interitemassociativity between list items are insufficient to supportdifferences in the recall of mixed lists in the second half ofthe list. While nonsignificant trends indicate better recall forHF than for LF words, it is clear that directional associativ-ity alone does not determine the relative success of recall inlate serial positions.

The correct serial recall data for pure lists revealed afrequency effect replicating previously reports (e.g., Hulmeet al., 2003, Experiment 2). The results for alternating listsindicated that lists beginning with HF items experience asmall advantage in correct recall, possibly as a consequenceof an item-specific contribution that operates at the start-of-listposition (Hulme et al., 2003, Experiment 2). Evidence for adirectionally sensitive contribution of interitem associativity

a bFig. 2 The conditionalprobabilities of continuous (a)and recovery (b) recall betweenconsecutive items for list typesof mixed list conditions.Dashed lines form the envelopeof the conditional probabilitiesof pure lists. Dashed circles inblack identify whether interitemassociativity predicts nodifference in likelihood of recall(serial positions 2–6). Solidcircles in black indicate whereeffects exist with Bonferroniadjustment (serial positions2–6). Solid circles in grayindicate marginal effects(serial positions 2–6)

Mem Cogn (2012) 40:1246–1256 1253

was produced in the recall of half lists, where recall mimickedpure lists across the first three serial positions. However, recallbeyond the halfway point of the list did not produce a reliablefrequency effect, except for the recency position, which mightalso respond to the item-specific properties of items.2

The possibility that interitem associativity between con-secutive list items had been obscured by the absolute levelsof recall of the previous items and masked in the second halfof the lists by instances of recovery after a failure to recallthe previous item was considered by fractionating the cor-rect recall data into continuous and recovery recall eventsand expressing these as conditional probabilities. Tests onthese measures, gauging how well the continuous data con-formed to a directional interitem associativity explanation,indicated that the first three to four list positions were wellaccounted for, although a significant difference in the re-covery data occurred for sequence lists, where recovery of HFwords in position 3 was greater than for LF words. Accord-ingly, the results of a second analysis reinforces the proposi-tion that while interitem effects of a directional nature operatefor the primacy portion of the list, these effects are lessinfluential for late list positions. Only one out of four positionswhere sequences of HF and LF items occurred in the secondhalf of the lists produced a significant effect, and this instancecoincided with the recency position.

This experiment did not find any order effects related toword frequency. If it is assumed that the proportion of ordererrors due to the loss of item information is the same for HFand LF words, this places the locus of frequency effect withdifferences in the retention of item information (Hulme et al.,

2003; Poirier & Saint-Aubin, 1996; Stuart & Hulme, 2000).The pattern of total item errors revealed that list compositioninfluenced the degree to which item recall was superior for HFwords; the effect was greater for pure than for mixed lists. Theinclusion of HF and LF words in the same list reduces errorsfor LF words and increases them for HF words, relative topure lists (Hulme et al., 2003). However, all formats producedfrequency effects in the item error data. Therefore, despitedifferences in item arrangement, LF words were associatedwith greater item error rates than were HF words across mixedlist conditions. These results contrast with those of Hulme etal. (2003, Experiment 2), who found a reversed frequencyeffect for item errors in alternating lists.

Accordingly, the first serial positions in recall are influ-enced by the preexperimental associations between adjacentlist items. Lists containing HF-to-LF transitions in the firstserial positions (e.g., alternating lists) will produce recallperformance consistent with the moderate strength of asso-ciation between items. Mixed lists that contain sequences ofHF items at the start of the list will be recalled reliably betterthan those containing sequences of LF words (half andsequence lists) in these serial positions. The similar levelsof recall observed for later list items of mixed lists might bea product of directional associativity in combination withthe level of available resources that are determined by theefficiency of output for earlier items. That is, for half andsequence lists, the absence of frequency effect might arisebecause lists beginning with HF items require fewerresources to output these early items than do pure LFlists and so, relatively speaking, the recall of LF items inhalf and sequence lists enjoy a benefit that buffers theeffects of directional associativity. Conversely, relative topure HF lists, half and sequence lists commencing withLF items might have fewer resources available for therecall of HF items later in the lists, with this cost coun-tering the facilitative effect of directional associativity.Furthermore, if item-to-item associativity was maintainedthroughout and obscured for final items, relationshipsbetween learning contexts and preexisting associationsof items in semantic memory, of the kind promoted bythe TCM (Howard & Kahana, 2002a, 2002b), might beresponsible.

While coarticulatory fluency is thought to be a minorfeature in the present investigation, it is worth consideringwhether the presumed effects of this variable (Woodward etal., 2008) are consistent with the patterns of results observedacross the conditions tested. A simple interpretation of thisapproach would predict that recall would be a function ofthe difficulty of coarticulating phonemes at word boundaries(as determined by frequency-based transitions) during re-hearsal. Consistent with this interpretation, the portion of thelist observed to reflect putative item-to-item associativeeffects does correspond to the subspan of items that are

2 While an effect at recency was consistent with item type in thisexperiment, Hulme et al. (2003) found a marginally significant nega-tive frequency effect—that is, better recall for LF than for HF words—in the recency position with alternating lists in their Experiment 2.

Table 1 Mean proportion of correct recall, order errors, and item errorsas a function of item type and list format

Correct Errors

Order Item

HF in pure lists .634 (.125) .090 (.044) .277 (.115)

LF in pure lists .434 (.101) .073 (.035) .494 (.095)

HF in alt. lists .559 (.128) .071 (.029) .370 (.119)

LF in alt. lists .524 (.126) .063 (.030) .413 (.110)

HF in half lists .592 (.113) .071 (.033) .337 (.101)

LF in half lists .525 (.136) .069 (.034) .406 (.124)

HF in seq. lists .613 (.107) .065 (.033) .322 (.087)

LF in seq. lists .554 (.135) .054 (.024) .391 (.120)

Note. Standard deviations are given in parentheses. HF, high frequen-cy; LF, low frequency; alt., alternating; seq., sequence

1254 Mem Cogn (2012) 40:1246–1256

cumulatively rehearsable within the interstimulus interval atpresentation (Page & Norris, 1998; Tan & Ward, 2008).3

Therefore, early sequences of HF items, argued to be morereproducible in terms of speech programming than sequen-ces of LF words, should be recalled better. However, theabsence of a frequency effect between sequences of HF andLF words in the second half of the list suggests that influ-ences wider than the coarticulation at word boundariesdetermine recall levels for these serial positions. It couldbe argued, for example, in the case of half lists, that thereduction in recall of HF items in the second half of the list,when compared with the recall of pure lists, would be due tothe impairment to speech planning brought about by therelative difficulty in sequencing three LF words in the firsthalf of the list. Similarly, the recall for LF items in thesecond half of the list, observed to be greater than the recallfor pure LF lists, could be explained as the benefit affordedto these items by the more efficient speech programming forthe first half of the list. One test of this proposal would be toconduct replications of these studies using articulatory sup-pression; however, Miller and Roodenrys (2012) found thatthe pattern of the frequency effect in half lists was notaltered by suppression. Therefore, if coarticulatory factorsplay a role in the formation of the effect, they must act at apoint prior to subvocal rehearsal. The perceptual–gesturalaccount of STM (Hughes, Marsh, & Jones, 2009; Jones,Hughes, & Macken, 2006; Murray & Jones, 2002), fromwhich the coarticulation account derives, argues that serialrecall does not involve mnemonic storage and processes.Instead, stimuli are organized into a perceptual stream that isused to sequence speech motor processes and produce anutterance for output. Accordingly, the coarticulation hypoth-esis would need to argue that some level of articulatoryplanning is the locus of the frequency effect.

The presence of preexperimental, item-to-item associa-tive effects in serial recall, as determined by word frequency,highlights yet another way in which the organization oflanguage knowledge constrains what is remembered overthe short term. The complexity of the relationship betweenlist composition, item arrangement, and the frequency effectis heightened by the possibility that directional associativityexists throughout recall but is masked for late serial positionsby the relative efficiencies or costs that arise from previousoutput. The findings reinforce the claim that the serial recalltask is not as simple as it may seem on the surface (Watkins &Watkins, 1977), and teasing apart the processes that contributeto serial recall at different points in the list poses a consider-able challenge.

Author Note Leonie M. Miller, School of Psychology, University ofWollongong; Steven Roodenrys, School of Psychology, University ofWollongong.

This research was conducted as part of the first author’s doctoralthesis and was funded by an Australian Postgraduate Award.

References

Allen, R., & Hulme, C. (2006). Speech and language processingmechanisms in verbal serial recall. Journal of Memory andLanguage, 55, 64–88.

Baayen, R. H., Piepenbrock, R., & Van Rijan, H. (1993). The CelexLexical Database. Philadelphia: University of Pennsylvania,Linguistic Data Consortium.

Baddeley, A. D. (1986). Working memory. New York: OxfordUniversity Press.

Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H.Bower (Ed.), The psychology of learning and motivation (Vol.8, pp. 47–90). New York: Academic Press.

Coltheart, M. (1981). The MRC Psycholinguistic Database. QuarterlyJournal of Experimental Psychology, 33A, 497–505. Retrievedfrom http://www.psy.uwa.edu.au/mrcdatabase/uwa_mrc.htm

DeCara, B., & Goswami, U. (2002). Similarity relations among spokenwords: The special status of rimes in English. Behavior ResearchMethods, Instruments, & Computers, 34, 416–423.

Deese, J. (1960). Frequency of usage and number of words in freerecall: The role of association. Psychological Reports, 7, 337–344.

Gregg, V., Freedman, C. M., & Smith, D. K. (1989). Word frequency,articulatory suppression and memory span. British Journal ofPsychology, 80, 363–374.

Henson, R. N. A., Norris, D. G., Page, M. P. A., & Baddeley, A. D.(1996). Unchained memory: Error patterns rule out chainingmodels of immediate serial recall. The Quarterly Journal ofExperimental Psychology, 49A, 80–115.

Howard, M. W., & Kahana, M. J. (2002a). A distributed representationof temporal context. Journal of Mathematical Psychology, 46,269–299.

Howard, M. W., & Kahana, M. J. (2002b). When does semanticsimilarity help episodic retrieval? Journal of Memory andLanguage, 46, 85–98.

Hughes, R. W., Marsh, J. E., & Jones, D. M. (2009). Perceptual–gestural (mis)mapping in serial short-term memory: The impactof talker variability. Journal of Experimental Psychology:Learning, Memory, and Cognition, 35, 1411–1425.

Hulme, C., Roodenrys, S., Schweickert, R., Brown, G. D. A., Martin,S., & Stuart, G. (1997). Word-frequency effects on short-termmemory tasks: Evidence for a redintegration process in immediateserial recall. Journal of Experimental Psychology: Learning,Memory, and Cognition, 23, 1217–1232.

Hulme, C., Stuart, G., Brown, G. D. A., & Morin, C. (2003). High- andlow-frequency words are recalled equally well in alternating lists:Evidence for associative effects in serial recall. Journal ofMemory and Language, 49, 500–518.

Jones, D. M., Hughes, R. W., & Macken, W. J. (2006). Perceptualorganization masquerading as phonological storage. Journal ofMemory and Language, 54, 265–281.

Larsen, R. J., & Marx, M. L., (1981). An introduction to mathematicalstatistics and its applications. Englewood Cliffs, NJ: Prentice‐Hall.

Miller, L. M., & Roodenrys, S. (2012). Serial recall, word frequencyand mixed lists: The influence of item arrangement. Journal ofExperimental Psychology: Learning, Memory, and Cognition.doi:10.1037/a0028470.

3 Although the presentation rate used in this experiment was one wordper second, all items had a CVC structure and so could be rehearsedrapidly.

Mem Cogn (2012) 40:1246–1256 1255

Morin, C., Poirier, M., Fortin, C., & Hulme, C. (2006). Word frequencyand the mixed-list paradox in immediate and delayed serial recall.Psychonomic Bulletin & Review, 13, 724–729.

Mueller, S. T., Seymour, T. L., Kieras, D. E., & Meyer, D. E. (2003).Theoretical implications of articulatory duration, phonologicalsimilarity, and phonological complexity in verbal working mem-ory. Journal of Experimental Psychology: Learning, Memory, andCognition, 29, 1353–1380.

Murdock, B. B., Jr. (1976). Item and order information in short-termserial memory. Journal of Experimental Psychology. General,105, 191–216.

Murray, A., & Jones, D. M. (2002). Articulatory complexity at itemboundaries in serial recall: The case of Welsh and English digitspan. Journal of Experimental Psychology: Learning, Memory,and Cognition, 28, 594–598.

Page, M. P. A., & Norris, D. (1998). The primacy model: A new modelof immediate serial recall. Psychological Review, 105, 761–781.

Poirier, M., & Saint-Aubin, J. (1996). Immediate serial recall, wordfrequency, item identity and item position. Canadian Journal ofExperimental Psychology, 50, 408–412.

Roodenrys, S., & Quinlan, P. T. (2000). The effects of stimulus set sizeand word frequency on verbal serial recall. Memory, 8, 71–78.

Roodenrys, S., Hulme, C., Lethbridge, A., Hinton, M., & Nimmo, L. M.(2002). Word-frequency and phonological-neighborhood effects onverbal short-term memory. Journal of Experimental Psychology:Learning, Memory, and Cognition, 28, 1019–1034.

Saint-Aubin, J., & LeBlanc, J. (2005). Word frequency effects inimmediate serial recall of pure and mixed lists: Tests of theassociative link hypothesis. Canadian Journal of ExperimentalPsychology, 59, 219–227.

Saint-Aubin, J., & Poirier, M. (1999). Semantic similarity and immediateserial recall: Is there a detrimental effect on order information?Quarterly Journal of Experimental Psychology, 52A, 367–394.

Saint-Aubin, J., & Poirier, M. (2005). Word frequency effects inimmediate serial recall: Item familiarity and item co-occurrencehave the same effect. Memory, 13, 325–332.

Schweickert, R. (1993). A multinomial processing tree model fordegradation and redintegration in immediate recall. Memory &Cognition, 21, 168–175.

Stuart, G., & Hulme, C. (2000). The effects of word co-occurrence onshort-term memory: Associative links in long-term memory affectshort-term memory performance. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 26, 796–802.

Tan, L., & Ward, G. (2008). Rehearsal in immediate serial recall.Psychonomic Bulletin & Review, 15, 535–542.

Tehan, G., &Humphreys, M. S. (1988). Articulatory loop explanations ofmemory span and pronunciation rate correspondences: A cautionarynote. Bulletin of the Psychonomic Society, 26, 293–296.

Toro, J. M., Nespor, M., Mehler, J., & Bonatti, L. L. (2008). Findingwords and rules in a speech stream: Functional differences betweenvowels and consonants. Psychological Science, 19, 137–144.

Watkins, O. C., & Watkins, M. J. (1977). Serial recall and the modalityeffect: Effects of word frequency. Journal of ExperimentalPsychology: Human Learning and Memory, 3, 712–718.

Woodward, A. J., Macken, W. J., & Jones, D. M. (2008). Linguisticfamiliarity in short-term memory: A role for (co-)articulatoryfluency? Journal of Memory and Language, 58, 48–65.

Wright, C. E. (1979). Duration differences between rare and commonwords and their implications for the interpretation of word frequencyeffects.Memory & Cognition, 7, 411–419.

1256 Mem Cogn (2012) 40:1246–1256

Conditional recall and the frequency effect in the serial recall task: an examination of...

Documents